Model for the Exceptional Reactivity of Peroxiredoxins 2 and 3 with Hydrogen Peroxide

Peroxiredoxins (Prx) are thiol peroxidases that exhibit exceptionally high reactivity toward peroxides, but the chemical basis for this is not well understood. We present strong experimental evidence that two highly conserved arginine residues play a vital role in this activity of human Prx2 and Prx3. Point mutation of either ArgI or ArgII (in Prx3 Arg-123 and Arg-146, which are ∼3–4 Å or ∼6–7 Å away from the active site peroxidative cysteine (Cp), respectively) in each case resulted in a 5 orders of magnitude loss in reactivity. A further 2 orders of magnitude decrease in the second-order rate constant was observed for the double arginine mutants of both isoforms, suggesting a cooperative function for these residues. Detailed ab initio theoretical calculations carried out with the high level G4 procedure suggest strong catalytic effects of H-bond-donating functional groups to the Cp sulfur and the reactive and leaving oxygens of the peroxide in a cooperative manner. Using a guanidinium cation in the calculations to mimic the functional group of arginine, we were able to locate two transition structures that indicate rate enhancements consistent with our experimentally observed rate constants. Our results provide strong evidence for a vital role of ArgI in activating the peroxide that also involves H-bonding to ArgII. This mechanism could explain the exceptional reactivity of peroxiredoxins toward H2O2 and may have wider implications for protein thiol reactivity toward peroxides.

reactive against peroxynitrite and other peroxides (4,6,7), but less so with other typical thiol-reactive reagents such as chloramines or alkylating electrophiles (5,8). There are six subfamilies (Prx1, Prx6, AhpE (one-cysteine peroxiredoxin from Mycobacterium tuberculosis), Prx5, Tpx, and bacterioferritin comigratory protein), categorized by amino acid sequence, which share a similar catalytic cycle. A number of highly conserved amino acid residues promote similar structures around their active site cysteine (peroxidatic cysteine; C p ) (9). Based on mechanistic considerations they are further subcategorized into 1-Cys and 2-Cys Prx. Prx2 and -3, the focus of this manuscript, are 2-Cys Prx that belong to the Prx1 subfamily.
The reactivity of the C p thiol toward H 2 O 2 is many orders of magnitude larger than that of free cysteine (e.g. the secondorder rate constants for the reaction of human Prx2 and Prx3 with H 2 O 2 are ϳ 10 7 M Ϫ1 s Ϫ1 (3,5), compared with ϳ1 M Ϫ1 s Ϫ1 for free cysteine or GSH (10)). Due to the low pK a Ͻ 6 of the C p sulfhydryl group, it is in its more nucleophilic thiolate form at physiological pH (5,11,12). However, this is not sufficient to explain the high reactivity of Prx with peroxides and further lowering of the pK a below 6 would decrease reactivity due to the linear free energy relationship (13). It is likely that conserved amino acid residues at the active site influence the rate of reaction via H-bonding interactions with the C p sulfur as well as with H 2 O 2 . Based on structural considerations, Hall et al. (9,14) proposed that a highly conserved arginine residue (ArgI) near the C p (Fig. 1, Prx3 Arg-123) activates the peroxide at the active site of Prx. There is another arginine (ArgII; Fig. 1a, Prx3 Arg-146) close to the active site (6 -7 Å from the C p sulfur, Fig. 1b) present in all members of the Prx1, Prx6, and AhpE subfamilies (see "Experimental Procedures"), which we hypothesize to have a role in the catalytic activity of these proteins.
Although an activator role for ArgI on H 2 O 2 has been proposed, only one study has tested the role of these arginine residues (along with other conserved amino acids) using site-directed mutagenesis. This showed that barley Prx ArgI and ArgII are important for its peroxidatic activity (15). However, activity was measured using a catalytic assay where the rate-determining step is reduction of the disulfide-linked Prx dimer, so it was not possible to assess the extent to which the reactivity of C p with H 2 O 2 was decreased. The effects of ArgI and ArgII on the kinetics of the reaction of C p with H 2 O 2 have not been investigated. In a combined crystallographic and quantum chemical investigation, Nakamura et al. (16) examined the archeal peroxiredoxin (ApTPx) in which a histidine residue (rather than an arginine) is in close proximity to C p . (This His residue is absent at the active site of mammalian Prx.) Their theoretical study focused on the reaction steps subsequent to the reduction of H 2 O 2 , and they concluded that a hypervalent sulfurane intermediate is involved in the formation at that stage of a sulfenic acid (16). In contrast, the theoretical component of the present investigation focuses on the initial catalytic reduction of H 2 O 2 by C p .
In the current study, we have mutated ArgI and ArgII in human Prx2 and Prx3 and measured their rates of reaction with H 2 O 2 . We demonstrate that both arginines play important roles in the reactivities of Prx2 and Prx3, independently from one another as well as in a cooperative fashion. We also used high level ab initio calculations to mechanistically explain our findings and propose a novel model to elucidate the high reactivity of peroxiredoxins with peroxides.

EXPERIMENTAL PROCEDURES
Reagents, preparation of recombinant Prx and site-directed mutagenesis of arginine residues, biophysical characterization of the recombinant Prx, computational details, and additional computational results are described in the supplemental "Methods." BLAST Search-To analyze whether Arg-146 of human Prx3 is conserved among other human peroxiredoxin isoforms and whether it is conserved in other organisms, a BLAST search against NCBI protein reference sequences was performed using residues 121-170 of human Prx3 as a query sequence (BLASTP 2.2.24). All of the 250 retrieved reference proteins aligning with the query sequence contained a corresponding arginine at a similar position in the protein.
Analysis of Oxidation State of Prx-Reduced monomer and disulfide-bonded dimer were separated by nonreducing SDS-PAGE on 15% gels as in Ref. 6. Gels were Coomassie Bluestained and scanned using a Fluor-S MultiImager (Bio-Rad). Bands were quantified using Quantity One software from Bio-Rad.
Kinetic Experiments-Prior to each experiment, Prx was reduced with either 25 mM DTT (Prx3) in 50 mM phosphate buffer for 2 h or by tris(2-carboxyethyl)phosphine (Prx2) for 1 h at pH 7.4. Buffer solutions contained 1 mM diethylenetriaminepentaacetic acid to chelate trace metal contaminants. Reduced Prx were separated from excess DTT using Micro Bio-Spin 6 columns (Bio-Rad,) prewashed with 200 l of 10 mg/ml bovine catalase and equilibrated with 50 mM phosphate buffer containing 100 M diethylenetriaminepentaacetic acid. Prx concentrations were measured using Bio-Rad DC Protein Assay Reagent and bovine serum albumin as standard. For kinetic analyses, the initial concentration of reduced Prx was calculated based on the relative intensity of the bands representing oxidized and reduced forms determined by gel electrophoresis.
Catalase Competition Assay-As described (  for the single and double mutants, respectively. Pseudo first-order rate constants were obtained by fitting the kinetic traces to a single exponential equation. Second, order rate constants were obtained by correcting the pseudo first-order rate constants with the applied H 2 O 2 concentrations. Kinetic curves were analyzed by SigmaPlot 11. Ab Initio Calculations-Full computational details are presented in the supplemental data. Here, we note in addition that our discussion is based on enthalpies (⌬H) rather than free energies (⌬G) because our small model approach would not adequately take into account the preorganization provided by the enzyme (18). However, the latter quantities show the same qualitative trends (supplemental Table S2). We refer to the electrophilic oxygen of H 2 O 2 (the one being attacked by HS Ϫ ) as O a and to the oxygen of the leaving hydroxide group as O b (14). The theoretical results reported in this paper correspond to the CPCM-G4//B3-LYP/6 -31ϩG(2df,p) level, which denotes the combination of a conductor-like polarizable continuum model (CPCM) solvation correction on top of a gas-phase G4//B3-LYP/6 -31ϩG(2df,p) enthalpy.

Two Highly Conserved Arginine Residues Play a Key Roles in the Peroxidase Activity of Prx2 and Prx3-Single substitution of
Arg-146 (ArgII) with various residues, Arg-123 (ArgI) with glycine, and a double substitution of both arginine residues with glycine, were introduced in Prx3. ArgI (Arg-127) and ArgII (Arg-150) in Prx2 were replaced with lysine, either singly or as a double mutant. WT recombinant proteins were constructed under the same conditions.
Commercially available and recombinant WT and mutant proteins displayed similar CD spectra, suggesting that secondary structures were not affected by mutations (supplemental Fig. S1). Melting points of the recombinant WT and mutant proteins (supplemental Table S1) show that, although R146A and R146H of Prx3 behaved differently, other mutants melted at a similar temperature to the WT.
The reactivities of the WT recombinant proteins with H 2 O 2 were investigated using competition kinetic assays against increasing amounts of bovine catalase. Prx2 and Prx3 are 2-Cys peroxiredoxins that form disulfide bonds between the C p on one chain and the resolving cysteine on the other (Cys-172 and Cys-168 in Prx2 and Prx3, respectively). Oxidation can therefore be followed as dimer formation on a nonreducing gel. As shown in Fig. 2a, addition of H 2 O 2 converted both Prx from reduced monomers to dimers. Catalase competed poorly against WT Prx2 and Prx3, with only marginal protection against dimerization at the highest concentration. This indicates a fast reaction between Prx and H 2 O 2 and agrees with previous studies using purified human erythrocyte Prx2 (5) and commercially available Prx3 (3).  Table 1 Prx also were competed against a constant amount of HRP for H 2 O 2 as described (12). H 2 O 2 reacts rapidly with the ferric form of HRP to give a stable intermediate (compound I) that has a lower Soret band absorbance at 403 nm. The reaction was quantified by measuring the decrease in this absorbance. The decrease was inhibited by increasing concentrations of the recombinant WT Prx (Fig. 2, c and d). Analysis of the competition data gave apparent second-order rate constants for reaction with H 2 O 2 that are similar to those reported for isolated and commercially available Prx (k eff ϳ 2 ϫ 10 7 M Ϫ1 s Ϫ1 ; Table 1) (3,5).
With the arginine mutants, catalase was fully protective even at the lowest concentration, and the Prx mutants did not inhibit the formation of compound I by HRP. This indicates that all of the arginine mutants are several orders of magnitude less reactive than the WT proteins. Reactions with these mutants are sufficiently slow to be followed over 15 min under pseudo firstorder conditions (excess of H 2 O 2 ), using gel electrophoresis to quantify the relative amounts of the monomeric (reduced) versus dimeric (oxidized) forms (Fig. 3). Dimer formation is a twostep process. The reaction of H 2 O 2 with the C p results in the formation of a sulfenic acid intermediate (Reaction 1), which reacts with the resolving cysteine of another protein molecule to give a disulfide-linked dimer (Reaction 2).

REACTIONS 1 and 2
Reactions were quenched with catalase or by alkylating the reactive thiol groups of the proteins with N-ethylmaleimide. Both methods gave similar rate constants, suggesting that Reaction 1 (between H 2 O 2 and C p ) was being measured, and not Reaction 2 (which is expected to be rapid) (19,20). A potential complication is that the Prx dimer bands on the gels represent both the single and double disulfide-bonded isoforms of the protein. Therefore, a proportionality constant needs to be incorporated into the rate law. This constant should have a value between 1 (if the reactivity of the dimer isoform with one disulfide bond is much greater than the monomer) and 2 (if the dimer is much less reactive). This introduces a 50% error in the apparent second-order rate constants for the reactions of H 2 O 2 with C p .
The kinetic traces that were obtained by following the loss of the monomeric forms fit well to single exponential equations, indicating a first-order dependence on Prx concentration (Fig.  3). The second-order rate constants (Table 1) were obtained by correcting the pseudo first-order rate constants with the H 2 O 2 concentrations. Second-order rate constants were similar at different H 2 O 2 concentrations (in the range of 13-50 M), suggesting overall second-order kinetics for the reaction. When ArgI was mutated, the second-order rate constants (for Prx3R123G and Prx2R127K) dropped by 5 orders of magnitude (to ϳ 10 2 M Ϫ1 s Ϫ1 ) compared with to WT. ArgI is therefore vital for the high reactivity of Prx.
Point mutation of ArgII (in Prx3R146A, Prx3R146H, Prx3R146K, Prx3R146G, and Prx2R150K) resulted in a similar decrease in the rate constant, demonstrating that ArgII is as important for the catalytic activity as ArgI. With both ArgI and ArgII mutated (Prx3R123G/R146G and Prx2R127K/R150K), the rate constants were a further 2 orders of magnitude smaller than for single mutants. This comprises a total of 7 orders of magnitude difference between double mutant and WT proteins (see Table 1), comparable with the difference in reactivity between free cysteine and C p . Ab Initio Calculations-High level calculations using the composite Gaussian-4 (G4) procedure were performed to assess possible H-bonding interactions at the active site that would lower the activation energy and increase the rate of the reaction of C p with H 2 O 2 . For reasons of computational efficiency, the C p and arginine residues were modeled by HS Ϫ and guanidinium cation, respectively. All of the uncatalyzed and catalyzed reductions of H 2 O 2 by HS Ϫ are found to proceed via an S N 2 mechanism. Fig. 4 and supplemental Fig. S2 display the reactant complexes, transition structures (TS), and product complexes located along the uncatalyzed and catalyzed reaction pathways. Table 2 gives the calculated reaction barriers (⌬H ‡ ) at 298 K.
In our analysis, the catalytic efficiency or rate enhancement of the catalyst is taken as the difference in barrier between the uncatalyzed process in aqueous solution and the enzymatic reaction (i.e. it is given by ⌬⌬H ‡ ϭ ⌬H ‡ (uncat,aq) Ϫ ⌬H ‡ (cat,enz) ), where the enzyme-like environment in the catalyzed reaction is simulated by a homogeneous polarizable continuum model with the dielectric constant ⑀ ϭ 4.0. We note that, according to the Arrhenius equation, a change of 5.7 kJ mol Ϫ1 in the barrier corresponds to a change of 1 order of magnitude in the reaction rate at 298 K.
Formation of the reactant complex (Fig. 4, Uncat-RC) is thermodynamically unfavorable in aqueous solution (⌬H (aq) ϭ ϩ5.2 kJ mol Ϫ1 ). Thus, the enthalpic barrier in aqueous solution is taken relative to the free reactants (⌬H ‡ (uncat,aq) ϭ 67.2 kJ mol Ϫ1 ). In all other cases, the barrier is taken as the energy of the TS relative to that of the reactant complex.
Single-arginine Model-To model the effects of ArgI, we incorporated a single guanidinium cation moiety into our calculations instead of HF. We located two reaction pathways,  Table S4).

and supplemental Fig. S2
Values are given as kJ mol Ϫ1 , calculated at the CPCM-G4//B3-LYP/6 -31ϩG(2df,p) level in a simulated protein-like phase.  Fig. 4), we find a barrier of just 26.4 kJ mol Ϫ1 ( Table 2). The catalytic efficiency is thus increased in this way to ⌬⌬H ‡ ϭ 40.8 kJ mol Ϫ1 , which corresponds to a rate enhancement of ϳ7 orders of magnitude relative to the uncatalyzed reaction in water. Consideration of weaker hydrogen-bond donors at S (namely H 2 O, NH 3 or H 2 CϭNH,

DISCUSSION
H 2 O 2 is a strong oxidant. However, its reactions with most thiols have a high activation barrier and therefore, although thermodynamically favored, are relatively slow. The structures of Prx overcome this barrier. This study investigated the roles of two highly conserved arginine residues (Fig. 1) in the exceptional reactivity of Prx with H 2 O 2 . Kinetic data obtained with various mutants of human Prx2 and Prx3 provided strong kinetic evidence that both arginine residues are vital for the peroxidase activities of these Prx. ArgI has been proposed to contribute significantly to the catalytic reaction. Indeed, we show mutations of ArgI in Prx2 and Prx3 result in a 5 orders of magnitude drop in reactivity with H 2 O 2 . Surprisingly, comparable losses in reactivity were also observed in ArgII mutants of both Prx. Importantly, simultaneous mutation of both arginine residues in Prx2 and Prx3 further decreased the rate constant to a total of 7 orders of magnitude less than the WT value.
High level quantum chemistry calculations provide mechanistic insight into the roles of these arginine residues in the reduction of H 2 O 2 by C p . The calculations indicate that partial protonation through H-bonding at the S, O a , and O b centers significantly lowers the activation barrier in a cooperative manner. A model using a single guanidinium cation moiety identifies two reaction pathways for catalysis ( Fig. 4) that lead to barrier reductions and are consistent with the position of ArgI at the Prx active site. The reductions in the barriers are 27.1 and 35.0 kJ mol Ϫ1 , respectively, compared with that for the uncatalyzed reaction. This translates to an increase of ϳ5-6 orders of magnitude in the second-order rate constant of the reaction. This is consistent with the experimentally observed difference in reactivity between the ArgI mutants and the WT proteins.
H-bonding of a guanidinium nitrogen to O a and either the leaving group (O b ) or the active site thiol has a profound effect on the catalytic activity. However, to account for the reactivity of the Prx compared with free cysteine, the barrier needs to be lowered further. This is achieved by introduction of a second,  Fig. 4). The barriers then become 26.4 and 27.8 kJ mol Ϫ1 , corresponding to a total of 7 orders of magnitude increase in rate constant, as observed experimentally for the double mutant proteins. Thus, the computational results as well as substantiating the proposal (14) that ArgI is crucial in the catalytic mechanism also indicate a critical role for a second protonating agent.
Computational modeling suggests that reduction of H 2 O 2 by cysteine proceeds via an S N 2 mechanism with a single transition structure (Scheme 1).
On the basis of our experimental observations and theoretical computations, we propose that ArgI is involved in lowering the electron density of the reacting oxygen (O a ) in H 2 O 2 (Fig. 1c) via H-bonding, making it a better electrophile. Simultaneously, ArgI could also be engaged in an H-bonding interaction with C p , which would not only contribute to lowering its pK a but would also anchor O a to C p (further increasing reactivity in line with theoretical calculations). This role is consistent with structural data on the position of ArgI in the Prx active site and similar to that proposed by Hall et al. (9,14).
Computational data indicate that additional interaction of either O b or C p with a strong H-donor in the Prx structure is needed to explain the high reactivity with H 2 O 2 . The dramatic decrease in reaction rate by ArgII mutation also needs to be taken into account. The crystal structures of Prx2 and Prx3 suggest that ArgII would be ideally located to donate an H-bond to O b . The dimensions of the active site (Fig. 1c) enable H 2 O 2 to sit within bonding distance of both ArgI and ArgII. This would lower the energy of the transition structure, thus facilitating the reaction. An H-bonding network between the two arginine residues and H 2 O 2 would also pull the two peroxide oxygens apart, thereby activating the oxidant by weakening the O a -O b bond. This would favor the OH ϩ transfer to C p as well as facilitate the departure of OH Ϫ .
This role for ArgII in the reactivity of Prx differs from that proposed by Hall et al. (9,14). They suggest that ArgII assists ArgI to adopt the correct conformation via an Arg-Glu-Arg H-bonding network to exert its catalytic effect. Our observation that point mutation of either arginine residue results in a similar drop in reactivity, for both Prx2 and Prx3, is consistent with this suggestion. However, the further 2 orders of magnitude SCHEME 1 drop in reactivity observed for double arginine mutants suggests a cooperative function for both residues.
We acknowledge that the orientation of H 2 O 2 in the transition structure of our proposed mechanism is different from that of the H 2 O 2 crystallized in the active site of the archeal peroxiredoxin ApTPx (21). Hall et al. (14) used the ApTpx-H 2 O 2 structure as the basis of their proposed model and also have shown that it is consistent with the orientation of different substrate mimics that have co-crystallized at the active site of other Prx. This model therefore favors involvement of residues other than ArgII in the H-bonding network. Further mechanistic and structural investigations are therefore required to reconcile these observations with our kinetic data.
In conclusion, our kinetic and computational studies substantiate structural studies and provide strong evidence for a vital role of ArgI in the high reactivity of Prx2 and Prx3 with H 2 O 2 . Our data also indicate that ArgII is critical in the activity of these proteins, and we propose that it could act through H-bonding to H 2 O 2 .